This paper is concerned with a chemotaxis system in a two-dimensional setting as follows:
{ut=Δu−χ∇⋅(u∇lnv)−κuv+ru−μu2+h1,vt=Δv−v+uv+h2,
with the parameters χ,κ,μ>0 and r∈R, and with the given functions h1,h2≥0. This model was originally introduced by Short et al for urban crime with the particular values χ=2,r=0 and μ=0, and the logistic source term ru−μu2 was incorporated into (⋆) by Heihoff to describe the fierce competition among criminals. Heihoff also proved that the initial-boundary value problem of (⋆) possesses a global generalized solution in the two-dimensional setting. The main purpose of this paper is to show that such a generalized solution becomes bounded and smooth at least eventually. In addition, the long-time asymptotic behavior of such a solution is discussed.
Citation: Zixuan Qiu, Bin Li. Eventual smoothness of generalized solutions to a singular chemotaxis system for urban crime in space dimension 2[J]. Electronic Research Archive, 2023, 31(6): 3218-3244. doi: 10.3934/era.2023163
[1] |
Jiayi Han, Changchun Liu .
Global existence for a two-species chemotaxis-Navier-Stokes system with |
[2] | Ling Xue, Min Zhang, Kun Zhao, Xiaoming Zheng . Controlled dynamics of a chemotaxis model with logarithmic sensitivity by physical boundary conditions. Electronic Research Archive, 2022, 30(12): 4530-4552. doi: 10.3934/era.2022230 |
[3] | Li Yang, Chunlai Mu, Shouming Zhou, Xinyu Tu . The global conservative solutions for the generalized camassa-holm equation. Electronic Research Archive, 2019, 27(0): 37-67. doi: 10.3934/era.2019009 |
[4] | Yunfei Yuan, Changchun Liu . Optimal control for the coupled chemotaxis-fluid models in two space dimensions. Electronic Research Archive, 2021, 29(6): 4269-4296. doi: 10.3934/era.2021085 |
[5] | Qihong Shi, Yaqian Jia, Xunyang Wang . Global solution in a weak energy class for Klein-Gordon-Schrödinger system. Electronic Research Archive, 2022, 30(2): 633-643. doi: 10.3934/era.2022033 |
[6] | Victor Ginting . An adjoint-based a posteriori analysis of numerical approximation of Richards equation. Electronic Research Archive, 2021, 29(5): 3405-3427. doi: 10.3934/era.2021045 |
[7] | Yin Yang, Sujuan Kang, Vasiliy I. Vasil'ev . The Jacobi spectral collocation method for fractional integro-differential equations with non-smooth solutions. Electronic Research Archive, 2020, 28(3): 1161-1189. doi: 10.3934/era.2020064 |
[8] | Changjian Wang, Jiayue Zhu . Global dynamics to a quasilinear chemotaxis system under some critical parameter conditions. Electronic Research Archive, 2024, 32(3): 2180-2202. doi: 10.3934/era.2024099 |
[9] | José Luis Díaz Palencia, Abraham Otero . Oscillatory solutions and smoothing of a higher-order p-Laplacian operator. Electronic Research Archive, 2022, 30(9): 3527-3547. doi: 10.3934/era.2022180 |
[10] | Ya Tian, Jing Luo . Boundedness and large time behavior of a signal-dependent motility system with nonlinear indirect signal production. Electronic Research Archive, 2024, 32(11): 6301-6319. doi: 10.3934/era.2024293 |
This paper is concerned with a chemotaxis system in a two-dimensional setting as follows:
{ut=Δu−χ∇⋅(u∇lnv)−κuv+ru−μu2+h1,vt=Δv−v+uv+h2,
with the parameters χ,κ,μ>0 and r∈R, and with the given functions h1,h2≥0. This model was originally introduced by Short et al for urban crime with the particular values χ=2,r=0 and μ=0, and the logistic source term ru−μu2 was incorporated into (⋆) by Heihoff to describe the fierce competition among criminals. Heihoff also proved that the initial-boundary value problem of (⋆) possesses a global generalized solution in the two-dimensional setting. The main purpose of this paper is to show that such a generalized solution becomes bounded and smooth at least eventually. In addition, the long-time asymptotic behavior of such a solution is discussed.
We study a class of logarithmic chemotaxis systems with the logistic source of the following form
{ut=Δu−χ∇⋅(u∇lnv)−κuv+ru−μu2+h1,vt=Δv−v+uv+h2, | (1.1) |
with the parameters χ,κ,μ>0 and r∈R. This model was proposed by Short et al. to describe the propagation of criminal activities with the particular values χ=2,r=0 and μ=0 ([1,2]), in which u(x,t) denotes the density of criminals, v(x,t) represents an abstract so-called attractiveness, the given function h1 denotes the density of additional criminals and h2 describes the source of attractiveness. The logistic source term, i.e., ru−μu2, is a fairly standard addition to chemotaxis models. Here, it was incorporated into the Short et al. model by Heihoff ([3]) to model the fierce competition among criminals for, e.g., good targets, which are limited resources. We refer to [4,5,6,7,8,9,10] for further developments of the Short et al. and to [11,12] for a review.
Mathematical analysis on (1.1) is still at quite an early stage and there are only a few relative results. For instance, for the Short et al. model, i.e., r=0 and μ=0, the local classical solution was obtained in [13], which is globally provided that either n=1 ([14,15]), or n≥2 and χ<2n ([16,17]), or the initial data and the given functions h1 and h2 are assumed to be small ([18,19]). As to the radial renormalized solvability, the global existence was established provided that either n=2 ([20]) or n=3 and χ∈(0,√3) ([21]); without requiring the symmetry hypothesis, the generalized solvability was obtained in [22] for any χ>0 and n=2. In addition, when Δu in the first equation in (1.1) is replaced by ∇⋅(∇um) with some m>0, the globally weak solvability was obtained in the two-dimensional setting provided that either m>32 ([23]) or m>1 and χ<√32 ([24]). We would like to remark that a reduced crime model, i.e., τut=Δu−χ∇⋅(u∇lnv) and vt=Δv−v+uv, admits an unbounded solution for appropriately large initial data, provided that n≥3, χ>0, and τ>0 is enough small ([25]). Finally, we mention there appear various studies on the variants of Short et al. model, see [26,27,28,29].
For the case of r∈R and μ>0, the corresponding initial-boundary value problem admits a generalized solution (in the sense of Definition 1.1 below) in the two-dimensional setting ([3]). To illustrate how critical the interaction between the term −μu2 in the first equation and the growth term +uv in the second equation is, the stronger logistic source, −μu2+α, with α>0 for n=2,3 ([3,30]) or α>n4−1 for n≥4 ([30]), was proved to be enough for the global existence of a classical solution. This also indicates that the regularity of the generalized solution structured in [3] is not enough to trigger a bootstrap argument to improve the regularity of such a solution, and thereby it is not known whether or not this generalized solution develops singularities. Therefore, motivated by [26,31,32,33,34], the main purpose of this paper is to reveal that the global generalized solution established in [3] at least eventually becomes bounded and smooth, and approaches spatial equilibria in the large time limit.
Precisely, we will present the eventual smoothness of the global generalized solution of the initial-boundary value problem:
{ut=Δu−χ∇⋅(u∇lnv)−κuv+ru−μu2+h1, x∈Ω, t>0,vt=Δv−v+uv+h2, x∈Ω, t>0,∇u⋅ν=∇v⋅ν=0, x∈∂Ω, t>0,u(x,0)=u0(x), v(x,0)=v0(x), x∈Ω, | (1.2) |
where ν denotes the exterior normal vector to the boundary ∂Ω and the initial data (u0,v0) fulfills that
{u0∈C0(¯Ω)withu0≥0andu0≢0,v0∈W1,∞(¯Ω)withinfx∈¯Ωv0>0. | (1.3) |
In order to specify the setup for our analysis, we assume throughout the sequel that
0≤hi∈C1(¯Ω×[0,∞))∩L∞(Ω×(0,∞)),i=1,2, | (1.4) |
with the additional properties that
inft>0∫Ωh2(x,t)dx>0, | (1.5) |
∫∞0∫Ωh1(⋅,t)dxds<∞, | (1.6) |
∫∞0∫Ω|h2(⋅,t)−h2,∞(⋅)|2dxds<∞ | (1.7) |
with some h2,∞∈C1(¯Ω).
Now, we briefly review the concept of generalized solution used in [3] for the initial-boundary value problem (1.2) as follows:
Definition 1.1. A pair of nonnegative functions (u,v) is called a global generalized solution to the initial-boundary value problem (1.2) if for any T>0,
1) it holds that for any q<∞
{v∈L∞(0,T;Lq(Ω)),lnv∈L2(0,T;W1,2(Ω)),u∈L2(Ω×(0,T))∩L∞(0,T;L1(Ω)),ln(1+u)∈L2(0,T;W1,2(Ω)),uv∈L1(Ω×(0,T)),v−1∈L∞(Ω×(0,T)); | (1.8) |
2) it holds that
∫Ωu(⋅,t)dx+∫t0∫Ω(κuv+μu2)dxds≤∫Ωu0dx+∫t0∫Ω(ru+h1)dxds,a.e.,in[0,T]; | (1.9) |
3) it holds that for 0≤φ(x,t)∈C∞0(¯Ω×[0,T)) with ∇φ⋅ν|∂Ω×(0,T)=0
−∫T0∫Ωln(u+1)φtdxdt−∫Ωln(u0+1)φ(⋅,0)dx≥∫T0∫Ωln(u+1)Δφdxdt+∫T0∫Ω|∇ln(u+1)|2φdxdt−χ∫T0∫Ωuu+1(∇ln(u+1)⋅∇lnv)φdxdt+χ∫T0∫Ωuu+1∇lnv⋅∇φdxdt−∫T0∫Ωκuvu+1φdxdt+∫T0∫Ωruu+1φdxdt−∫T0∫Ωμu2u+1φdxdt+∫T0∫Ωh1φdxdt; | (1.10) |
4) it holds that for all φ∈L∞(0,T;Lq(Ω))∩L2(0,T;W1,2(Ω)) with φt∈L2(Ω×(0,T)), compact support in ¯Ω×[0,T) and q<∞
∫T0∫Ωvφtdxdt+∫Ωv0φ(⋅,0)dx=∫T0∫Ω∇v⋅∇φdxdt+∫T0∫Ωvφdxdt−∫T0∫Ωuvφdxdt−∫T0∫Ωh2φdxdt. | (1.11) |
With Definition 1.1 at hand, letting
η:=min{infx∈Ωv0(x)e−1,14πe−1−(diamΩ)22{infs>0∫Ωh2(⋅,s)dx}}, | (1.12) |
our main results read as follows.
Theorem 1.1. Assume that (1.3)–(1.7) hold. Let κ,χ,μ>0, r∈R and Ω⊂R2 be a bounded convex domain with smooth boundary, and let (u,v) be a generalized solution of (1.2) in the sense of Definition 1.1. Under the additional assumption that r<κη with η determined by (1.12), there exists t0>0, with the properties that u(x,t)≥0 and v(x,t)>0 for any x∈¯Ω and any t≥t0, and
u∈C2,1(¯Ω×[t0,∞)),v∈C2,1(¯Ω×[t0,∞)), | (1.13) |
and that (u,v) solves the initial-boundary value problem (1.2) classically in Ω×(t0,∞). Moreover, (u,v) fulfills that
‖u(⋅,t)‖L∞+‖v(⋅,t)−v∞(⋅)‖L∞→0,ast→∞, | (1.14) |
where v∞ denotes the solution of the boundary value problem
{−Δv∞+v∞=h2,∞,x∈Ω,∇v∞⋅ν=0,x∈∂Ω. | (1.15) |
Technical strategy and structure of the article
The objective of this paper, motivated by [26,31,32,33,34], is to present that the global generalized solution of the initial-boundary value problem (1.2) at least eventually becomes bounded and smooth, and approaches spatial equilibria in a large time limit. To this end, the key steps are to establish a series of uniform a-priori estimates, in which the starting point is to get the uniform-in-(ε,t) lower bound for vε, see Lemma 2.1. We would like to remark that, for the linear signal production mechanism the combinational functional of the form
∫Ωuεlnuε+12|∇^vε|2+1edx |
where ^vε:=vε−v∞ and v∞ is a classical solution to the boundary value problem (1.15), is usually adopted to get the desired a-priori estimates (e.g., [35]). However, thanks to the presence of the nonlinear signal production mechanism, such functional is invalid for our case. Here, our novelty of the analysis consists of tracking the time evolution of the combinational functional of the form
∫Ωbuε+12u2ε+12|∇^vε|2dx,t≥T0 |
with some waiting time T0 and some b>0, see Lemmas 3.4 and 3.5. From this, the key L2-bound of uε is obtained, and an application of the standard bootstrap techniques shows that the generalized solution established in [3] becomes bounded and smooth at least eventually.
The rest of this paper is arranged as follows. Some preliminaries are given in Section 2. A-priori estimates are established in Section 3. Section 4 is devoted to showing the eventual smoothness, and the last section presents the large-time behavior desired in Theorem 1.1.
A generalized solution of the initial-boundary value problem (1.2) can be obtained by an approximation procedure ([3,22]). Accordingly, we shall consider the following approximate problem
{uεt=Δuε−χ∇⋅(uε∇lnvε)−κuεvε+ruε−μu2ε+h1,x∈Ω,t>0,vεt=Δvε−vε+uεvε1+εuεvε+h2,x∈Ω,t>0,∂uε∂ν=∂vε∂ν=0,x∈∂Ω,t>0,uε(x,0)=u0(x),vε(x,0)=v0(x),x∈Ω. | (2.1) |
An application of the strategy invoking the contraction mapping principle and the well-known pointwise positivity property of the Neumann heat semigroup, as in [13,16,36,37], ensures the global existence of the classical solution to the approximate problems (2.1).
Lemma 2.1. Let the assumptions (1.3)–(1.4) hold. For each ε∈(0,1), there exists a unique pair (uε,vε) of positive functions, with the properties that for any T>0 and ι>2
{uε∈C0(¯Ω×[0,T])∩C2,1(¯Ω×(0,T]),vε∈C0(0,T;W1,ι(¯Ω))∩C2,1(¯Ω×(0,T]), |
such that (uε,vε) solves the approximate problem (2.1) classically in Ω×[0,∞).
Proof. By a slight adaptation of the proof of [3,Lemma 2.3] (see also [22]), we can easily get the desired results.
Note that thanks to the non-negativity of (uε,h2) and the variation-of-constants formula for vε, namely,
vε(⋅,t)=et(Δ−1)v0+∫t0e(t−s)(Δ−1)(uεvε1+εuεvε+h2)(⋅,s)ds, | (2.2) |
it is clear that
vε(⋅,t)≥et(Δ−1)v0≥e−tinfx∈Ωv0(x),t>0, | (2.3) |
which is adequate for establishing the global existence of generalized solutions, see [3]. However, to get eventual smoothness of generalized solutions, the uniform-in-t lower bound for vε will be necessary.
Lemma 2.2. Let Ω⊂R2 be a bounded convex domain with smooth boundary and (1.3)–(1.7) hold. Then we have
vε(⋅,t)≥η,t>0, | (2.4) |
where η is determined by (1.12).
Proof. Thanks to (2.3), we have
vε(⋅,t)≥e−1infx∈Ωv0(x)for allt≤1. | (2.5) |
For t>1, due to the convexity of Ω, the well-known pointwise positivity property of the Neumann heat semigroup ensures that
etΔf≥14πte−(diamΩ)24t∫Ωfdx,t>0, |
where f∈C0(¯Ω) (cf. [38,Lemma 2.3] and [39,Lemma 3.1]), which, combined with (2.2), (1.5) and the non-negativity of (uε,v0), implies that
vε(⋅,t)≥∫t−120e(t−s)(Δ−1)h2(⋅,s)ds≥∫t−120e−(t−s)14π(t−s)e−(diamΩ)24(t−s)∫Ωh2(⋅,s)dxds for allt>1. |
It follows that for t>1
vε(⋅,t)≥14π{infs>0∫Ωh2(⋅,s)dx}∫t12e−ss−1e−(diamΩ)24sds≥14π{infs>0∫Ωh2(⋅,s)dx}∫112e−ss−1e−(diamΩ)24sds. |
Based on this, we further get that
vε(⋅,t)≥14π{infs>0∫Ωh2(⋅,s)dx}⋅e−1e−(diamΩ)22. |
This, together with (2.5), entails the desired (2.4).
Next, we are concerned with the decay in a linear differential inequality, which is an extended version of [22,Lemma 2.5].
Lemma 2.3. Let ε∈(0,1), yε∈C1([0,∞)) be non-negative functions satisfying
yε(0)=m | (2.6) |
with some positive constant m independent of ε. If there exist a positive constant k and a nonnegative function gε(t)∈C([0,∞))∩L∞([0,∞)) which satisfies
limt→∞∫t+1tgε(s)ds=0 uniformly inε, | (2.7) |
‖gε‖L∞(0,∞)≤μforsomeμindependentof(ε,t), | (2.8) |
such that for each ε>0,
y′ε(t)+kyε(t)≤gε(t) for allt>0, | (2.9) |
then
yε(t)→0 ast→∞ uniformly inε. | (2.10) |
At the end of this section, we recall the result on the solvability of the boundary value (1.15), which directly follows from [40].
Lemma 2.4. For given h2,∞∈C1(¯Ω), the problem (1.15) possesses a unique classical solution v∞ fulfilling that v∞∈C2+θ(¯Ω) for some θ∈(0,1).
A straightforward consequence of Lemma 2.1 is the following L1-decay on the component uε.
Lemma 3.1. Let all assumptions in Theorem 1.1 be fulfilled. Then there exists C>0, independent of (ε,t), such that
∫Ωvε(⋅,t)dx+∫Ωuε(⋅,t)dx+∫t0∫Ωuε(⋅,s)vε(⋅,s)dxds+∫t0∫Ωu2ε(⋅,s)dxds≤C,t>0, | (3.1) |
and that
∫Ωuε(⋅,t)dx→0ast→∞uniformlyinε, | (3.2) |
∫t+1t∫Ωuε(⋅,s)vε(⋅,s)dxds+∫t+1t∫Ωu2ε(⋅,s)dxds→0ast→∞uniformlyinε. | (3.3) |
Proof. Invoking (2.4) and taking c1∈(0,κ), we obtain
ddt∫Ωuεdx+(κ−c1)η∫Ωuεdx+c1∫Ωuεvεdx+μ∫Ωu2εdx≤r∫Ωuεdx+∫Ωh1dx. |
Under the assumption that r<κη, we can further take c1 sufficiently close to 0 such that
c2:=(κ−c1)η−r>0, |
and thereby get
ddt∫Ωuεdx+c2∫Ωuεdx+c1∫Ωuεvεdx+μ∫Ωu2εdx≤∫Ωh1dx. | (3.4) |
We now integrate the second equation in (2.1) over Ω to obtain
ddt∫Ωvεdx+∫Ωvεdx=∫Ωuεvεdx+∫Ωh2dx, |
which, together with (3.4), ensures
ddt{∫Ωuεdx+c1∫Ωvεdx}+c2∫Ωuεdx+c1∫Ωvεdx+μ∫Ωu2εdx≤∫Ωh1dx+∫Ωh2dx. |
Setting y(t):=∫Ωuεdx+c1∫Ωvεdx and c3:=min{c2,1}, it follows from (1.4) that
y′(t)+c3y(t)≤c4:=‖h1‖L∞(Ω×(0,∞))|Ω|+‖h2‖L∞(Ω×(0,∞))|Ω|. |
A standard ODE technique shows that
∫Ωuεdx+c1∫Ωvεdx≤c5:=max{∫Ωu0dx+c1∫Ωv0dx,c4c3},t>0. | (3.5) |
On the other hand, integrating (3.4) over [0,t], for any t>0 we infer that
∫Ωuεdx+c2∫t0∫Ωuεdxds+c1∫t0∫Ωuεvεdxds+μ∫t0∫Ωu2εdxds≤∫Ωu0dx+∫t0∫Ωh1dx, |
which, with the help of (1.6) and (3.5), ensures (3.1).
Moreover, thanks to (1.6) and (1.4), it follows that
∫t+1t∫Ωh1dxds→0,ast→∞, |
which, together with Lemma 2.3 and (3.4), entails that the decay (3.2) holds as desired. Integrating (3.4) over [t,t+1], for any t>0 we have
∫Ωuε(⋅,t+1)dx+c1∫t+1t∫Ωuεvεdxds+μ∫t+1t∫Ωu2εdxds≤∫Ωuε(⋅,t)dx+∫t+1t∫Ωh1dxds. |
Recalling (1.6) and (3.2), we arrive at (3.3).
To proceed further, we track the time evolution of ‖vε(⋅,t)−v∞(⋅)‖L2, where v∞ is classical solution of (1.15). For convenience, we set ˆvε:=vε−v∞. Thanks to (1.15) and (2.1), for (uε,vε) given in Lemma 2.1, the initial-boundary value problem
{ˆvεt=Δˆvε−ˆvε+uεvε1+εuεvε+h2−h2,∞,x∈Ω,t>0,∇ˆvε⋅ν=0,x∈∂Ω,t>0,ˆvε(x,0)=v0(x)−v∞(x),x∈Ω | (3.6) |
admits a unique classical solution ˆvε.
Lemma 3.2. Let all assumptions in Theorem 1.1 be in force. Then there exists C>0, independent of (ε,t), such that
‖^vε(⋅,t)‖2L2≤C,t>0 | (3.7) |
and
∫t0∫Ω|∇^vε|2dxds+∫t0∫Ω|^vε|2dxds≤C,t>0. | (3.8) |
Proof. Testing the first equation of (3.6) with ^vε, yields
12ddt∫Ω|^vε|2dx≤−∫Ω|∇^vε|2dx−∫Ω|^vε|2dx+∫Ω^vεuεvε1+εuεvεdx+∫Ω(h2−h2,∞)^vεdx,t>0. |
Using Hölder's inequality and recalling the definition of ^vε, we have
∫Ω^vεuεvε1+εuεvεdx=∫Ω^vε2uε1+εuεvεdx+∫Ωv∞uε^vε1+εuεvεdx≤‖uε‖L2‖^vε‖2L4+‖v∞‖L∞‖uε‖L2‖^vε‖L2. |
An application of the Gagliardo-Nirenberg inequality and Young's inequality implies that
‖uε‖L2‖^vε‖2L4≤c1‖uε‖L2(‖^vε‖L2‖∇^vε‖L2+‖^vε‖2L2)≤14‖∇^vε‖2L2+c2‖uε‖2L2‖^vε‖2L2+14‖^vε‖2L2. | (3.9) |
In addition, we have
∫Ω(h2−h2,∞)^vεdx≤14‖^vε‖2L2+∫Ω|h2−h2,∞|2dx,‖v∞‖L∞‖uε‖L2‖^vε‖L2≤14‖^vε‖2L2+‖v∞‖2L∞‖uε‖2L2. |
Collecting these, we arrive at
12ddt∫Ω|^vε|2dx+34∫Ω|∇^vε|2dx+14∫Ω|^vε|2dx≤c2‖uε‖2L2‖^vε‖2L2+‖v∞‖2L∞‖uε‖2L2+∫Ω|h2−h2,∞|2dx,t>0. | (3.10) |
Setting y(t)=‖^vε(⋅,t)‖2L2 and a(t)=‖uε(⋅,t)‖2L2, it follows that
y′(t)+12y(t)≤2c2a(t)y(t)+b(t),b(t):=2‖v∞‖2L∞‖uε‖2L2+2∫Ω|h2−h2,∞|2dx. |
A standard ODE technique shows
y(t)≤y(0)e2c2∫t0a(s)ds−12t+e2c2∫t0a(s)ds−12t∫t0b(s)e−2c2∫s0a(τ)dτ+12sds. |
Note that, thanks to (3.1), there exists c3>0, independent of (ε,t), such that ∫t0a(s)ds≤c3. Hence, we arrive at
y(t)≤c3y(0)e−12t+c3e−12t∫t0b(s)e12sds. |
Using (1.7) and (3.1) again, there exists c4>0, independent of (ε,t), such that
c3e−12t∫t0b(s)e12sds≤c3∫t0b(s)ds≤c4,t>0. |
Hence, there exists C>0, independent of (ε,t), such that (3.7) holds. Moreover, thanks to (3.10) we can find c5>0, independent of (ε,t), such that
12ddt∫Ω|^vε|2dx+34∫Ω|∇^vε|2dx+14∫Ω|^vε|2dx≤c5‖uε‖2L2+∫Ω|h2−h2,∞|2dx,t>0. | (3.11) |
We now integrate this equation over [0,t] to get
12∫Ω|^vε|2dx+34∫t0∫Ω|∇^vε|2dxds+14∫t0∫Ω|^vε|2dxds≤12‖v0−v∞‖2L2+c5∫t0‖uε‖2L2ds+∫t0∫Ω|h2−h2,∞|2dxds, |
which, combined with (3.1) and (1.7), gives us the desired (3.8).
We would like to remark that although we have obtained (3.11) and can infer from (3.3) and (1.7) that for any ε∈(0,1)
∫t+1t‖uε‖2L2ds+∫t+1t∫Ω|h2−h2,∞|2dxds→0,ast→∞, | (3.12) |
we cannot directly get the desired decay on ‖^vε(⋅,t)‖L2 by Lemma 2.3 due to the absence of the bound of ‖uε‖L∞(t,t+1;L2). Here, compared with (3.2), we need a new method to get decay on ^vε.
Lemma 3.3. Let all assumptions in Theorem 1.1 be in force. Then
∫Ω|^vε|2(⋅,t)dx→0 ast→∞ uniformly inε, | (3.13) |
∫t+1t∫Ω|∇^vε|2dxds→0 ast→∞ uniformly inε. | (3.14) |
Proof. In fact, integrating (3.11) over [t,t+1] yields that
12∫Ω|^vε|2(⋅,t+1)dx−12∫Ω|^vε|2(⋅,t)dx+34∫t+1t∫Ω|∇^vε|2dxds+14∫t+1t∫Ω|^vε|2dxds≤c5∫t+1t‖uε‖2L2ds+∫t+1t∫Ω|h2−h2,∞|2dxds. | (3.15) |
By setting z(t):=12∫t+1t∫Ω|^vε|2dxds, we have
z′(t)+12z(t)≤c5∫t+1t‖uε‖2L2ds+∫t+1t∫Ω|h2−h2,∞|2dxds. |
We now infer from (1.4), (1.7) and (3.1) that there exists C>0, independent of (ε,t), such that
gε:(t)=∫t+1t‖uε‖2L2ds+∫t+1t∫Ω|h2−h2,∞|2dxds≤C,t>0, |
which, combined with Lemma 2.3, ensures that
z(t):=12∫t+1t∫Ω|^vε|2dxds→0 ast→∞ uniformly inε. | (3.16) |
On the other hand, letting y(t):=∫Ω{(1+μ−1‖v∞‖2L∞)uε(⋅,t)+12|^vε|2(⋅,t)}dx we can infer from (3.4) and (3.10) that there exist ci, i=1,2,3, independent of (ε,t), such that for any t>0
y′(t)+c1y(t)+34∫Ω|∇^vε|2dx+(μ−c2‖^vε‖2L2)∫Ωu2εdx≤c3∫Ωh1+|h2−h2,∞|2dx. | (3.17) |
From (3.16), (3.2) and the assumptions (1.6) and (1.7), there must exist T∗ large enough, independent of ε, such that
12∫T∗+1T∗‖^vε(⋅,t)‖2L2dt+c3∫∞T∗∫Ωh1+|h2−h2,∞|2dxds≤μ16c2,(1+μ−1‖v∞‖2L∞)∫Ωuε(⋅,t)dx≤μ16c2,t≥T∗, |
by which the mean value theorem implies there exists ˆt0∈(T∗,T∗+1), depending on ε, such that
(1+μ−1‖v∞‖2L∞)∫Ωuε(⋅,ˆt0)dx+12‖^vε(⋅,ˆt0)‖2L2+c3∫∞ˆt0∫Ωh1+|h2−h2,∞|2dxds≤μ8c2. | (3.18) |
Invoking these, we can claim that
‖^vε(⋅,t)‖2L2≤μ2c2,t≥ˆt0. | (3.19) |
In fact, the continuity of ‖^vε(⋅,t)‖2L2, combined with (3.18), ensures that
˜T:=sup{t|supˆt0≤s≤t‖^vε(⋅,t)‖2L2≤μ2c2}>ˆt0, | (3.20) |
and so we only need to show that ˜T=∞. If on the contrary, there must hold
supˆt0≤s≤˜T‖^vε(⋅,t)‖2L2=μ2c2. | (3.21) |
However, it follows from (3.17) and (3.20) that for t∈[ˆt0,˜T]
y′(t)+c1y(t)+34∫Ω|∇^vε|2dx+12μ∫Ωu2εdx≤c3∫Ωh1+|h2−h2,∞|2dx. |
By employing the standard ODE techniques, we arrive at
y(t)≤e−c1(t−ˆt0)y(ˆt0)+c3e−c1t∫tˆt0ec1s∫Ωh1+|h2−h2,∞|2dxds≤y(ˆt0)+c3∫tˆt0∫Ωh1+|h2−h2,∞|2dxds, |
which, with the help of (3.18), ensures
y(t)≤μ8c2,t∈[ˆt0,˜T]. |
Recalling the definition of y(t), we have
‖^vε(⋅,t)‖2L2≤μ4c2,t∈[ˆt0,˜T], |
which contradicts (3.21). Thus we have that ˜T=∞, and prove (3.19) as desired.
Thanks to the validity of (3.19), it follows from (3.17) that
y′(t)+c1y(t)+34∫Ω|∇^vε|2dx+μ2∫Ωu2εdx≤c3∫Ωh1+|h2−h2,∞|2dx,t≥ˆt0. | (3.22) |
Based on the assumptions (1.4), (1.6), (1.7) and Lemma 2.3, (3.22) ensures
y(t)→0 ast→∞ uniformly inε, | (3.23) |
which is enough for (3.13) by recalling the definition of y(t).
To get (3.14), integrating (3.22) over [t,t+1], yields
y(t+1)+34∫t+1t∫Ω|∇^vε|2dxds+μ2∫t+1t∫Ωu2εdxds≤y(t)+c3∫t+1t∫Ωh1+|h2−h2,∞|2dxds, |
which, combined with (3.23), (1.6) and (1.7) again, entails that (3.14) holds as desired.
In the sequel, we will use (3.2), (3.3) and (3.14) to obtain the uniform in ε bound for the entropy functional, denoted by
Eε(t):=12∫Ωu2ε+|∇^vε|2dx,t>0. | (3.24) |
To achieve it, we first manage to achieve the following estimate.
Lemma 3.4. Let all assumptions in Theorem 1.1 hold. Then there exist a1,a2,a3>0, such that for any ε∈(0,1),
E′(t)+12∫Ω|∇uε|2dx+κη−r2∫Ωu2εdx+μ∫Ωu3εdx+(12−a1‖uε‖2L2(‖∇^vε‖2L2+1))∫Ω|Δ^vε|2dx+(1−a2‖uε‖2L2)∫Ω|∇^vε|2dx≤a3{‖h1‖L1+‖uε‖2L2+‖h2−h2,∞‖2L2},t>0, | (3.25) |
where η is given by (2.4).
Proof. Invoking integration by parts, we have
12ddt∫Ωu2εdx=∫Ωuε(Δuε−χ∇⋅(uε∇lnvε)−κuεvε+ruε−μu2ε+h1)dx=−∫Ω|∇uε|2dx+χ∫Ω∇uε⋅(uε∇lnvε)dx−κ∫Ωu2εvεdx+r∫Ωu2εdx−μ∫Ωu3εdx+∫Ωh1uεdx=:P1+P2+P3+P4+P5+P6. |
Since r<κη (η given in (2.4)), it follows that c1:=κη−r>0, and thereby implies from (2.4) that
P3+P4≤−κη∫Ωu2εdx+r∫Ωu2εdx=−c1∫Ωu2εdx. |
And using Young's inequality yields
P6≤c12∫Ωu2εdx+c2∫Ωh21dx. |
For P2, Hölder's inequality and (2.4) imply
P2≤χη−1‖∇uε‖L2‖uε∇vε‖L2≤χη−1‖∇uε‖L2‖uε‖L4‖∇vε‖L4, |
which, together with Young's inequality, entails
P2≤14‖∇uε‖2L2+c3‖uε‖2L4‖∇vε‖2L4. |
Recalling the Gagliardo-Nirenberg inequality
‖f‖2L4≤c4(‖f‖L2‖∇f‖L2+‖f‖2L2), |
we get
‖uε‖2L4≤c5(‖uε‖L2‖∇uε‖L2+‖uε‖2L2), |
and infer from the elliptic estimates that
‖∇vε‖2L4≤c6‖∇vε‖L2‖∇vε‖H1≤c7‖∇vε‖L2‖Δvε‖L2. |
In view of these, we arrive at
‖uε‖2L4‖∇vε‖2L4≤c8(‖uε‖L2‖∇uε‖L2+‖uε‖2L2)‖∇vε‖L2‖Δvε‖L2≤14‖∇uε‖2L2+c28‖uε‖2L2‖∇vε‖2L2‖Δvε‖2L2+c8‖uε‖2L2‖∇vε‖L2‖Δvε‖L2. |
Collecting these and using Young's inequality, we have
12ddt∫Ωu2εdx+12∫Ω|∇uε|2dx+c12∫Ωu2εdx+μ∫Ωu3εdx≤c2∫Ωh21dx+c28‖uε‖2L2‖∇vε‖2L2‖Δvε‖2L2+c8‖uε‖2L2‖∇vε‖L2‖Δvε‖L2≤c2∫Ωh21dx+2c28‖uε‖2L2‖∇vε‖2L2‖Δvε‖2L2+c9‖uε‖2L2. |
Recalling ^vε:=vε−v∞ and invoking Lemma 2.4, it follows that
‖∇vε‖2L2‖Δvε‖2L2≤4(‖∇^vε‖2L2+‖∇v∞‖2L2)(‖Δ^vε‖2L2+‖Δv∞‖2L2)≤c10(‖∇^vε‖2L2‖Δ^vε‖2L2+‖∇^vε‖2L2+‖Δ^vε‖2L2+1). |
This leads to
12ddt∫Ωu2εdx+12∫Ω|∇uε|2dx+c12∫Ωu2εdx+μ∫Ωu3εdx≤c2∫Ωh21dx+c11‖uε‖2L2(‖∇^vε‖2L2+1)‖Δ^vε‖2L2+c12‖uε‖2L2‖∇^vε‖2L2+c13‖uε‖2L2. | (3.26) |
On the other hand, we can test the first equation in (3.6) with −Δ^vε to get
12ddt∫Ω|∇^vε|2dx+∫Ω|Δ^vε|2dx+∫Ω|∇^vε|2dx=∫Ωuεvε1+εuεvε(−Δ^vε)dx+∫Ω(h2−h2,∞)(−Δ^vε)dx, |
which, with the help of Young's inequality, shows
12ddt∫Ω|∇^vε|2dx+12∫Ω|Δ^vε|2dx+∫Ω|∇^vε|2dx≤∫Ωu2εv2εdx+∫Ω|h2−h2,∞|2dx. |
Hölder's inequality, combined with the Gagliardo-Nirenberg inequality and the elliptic estimates, entails
∫Ωu2εv2εdx≤‖uε‖2L2‖vε‖2L∞≤c14‖uε‖2L2(‖vε‖L2‖Δvε‖L2+‖vε‖2L2), |
which, based on (3.7), Lemma 2.4 and the fact that ^vε:=vε−v∞, leads to
∫Ωu2εv2εdx≤c14‖uε‖2L2((‖^vε‖L2+‖v∞‖L2)(‖Δ^vε‖L2+‖Δv∞‖L2)+(‖^vε‖L2+‖v∞‖L2)2)≤c15‖uε‖2L2(‖Δ^vε‖L2+1). |
In the light of Young's inequality, it follows that
∫Ωu2εv2εdx≤c16‖uε‖2L2(‖Δ^vε‖2L2+1). |
Hence, we arrive at
12ddt∫Ω|∇^vε|2dx+12∫Ω|Δ^vε|2dx+∫Ω|∇^vε|2dx≤c16‖uε‖2L2(‖Δ^vε‖2L2+1)+∫Ω|h2−h2,∞|2dx, |
which, together with (3.26), ensures that
E′(t)+12∫Ω|∇uε|2dx+c12∫Ωu2εdx+μ∫Ωu3εdx+12∫Ω|Δ^vε|2dx+∫Ω|∇^vε|2dx≤c2∫Ωh21dx+c11‖uε‖2L2(‖∇^vε‖2L2+1)‖Δ^vε‖2L2+c12‖uε‖2L2‖∇^vε‖2L2+c13‖uε‖2L2+c16‖uε‖2L2(‖Δ^vε‖2L2+1)+∫Ω|h2−h2,∞|2dx. |
Note that due to (1.4), we have
∫Ωh21dx≤‖h1‖2L∞(Ω×(0,∞))|Ω|. |
Hence, collecting these and recalling the definition of c1, we can get the validity of (3.25).
The uniform convergence properties previously established in Lemmas 3.1 and 3.3, combined with a continuation argument, are enough to show that there exists T0 large enough such that the variable coefficient in (3.25) maintains nonnegativity whenever t≥T0, which shall eventually lead to the following crucial estimates.
Lemma 3.5. There exist T0 large enough and a4>0, independent of ε, such that for any ε∈(0,1)
∫Ωu2ε(⋅,t)dx+∫Ω|∇^vε|2(⋅,t)dx≤a4,t≥T0, | (3.27) |
∫ts∫Ω|∇uε|2dxdτ+∫ts∫Ω|Δ^vε|2dxdτ≤a4,t≥s≥T0. | (3.28) |
Proof. Combining with (3.4) and (3.25), and setting y(t):=a3μ∫Ωuεdx+Eε(t), there exist c1>0 and c2>0, independent of (ε,t), such that
y′(t)+c1y(t)+(12−a1‖uε‖2L2(‖∇^vε‖2L2+1))∫Ω|Δ^vε|2dx+12∫Ω|∇uε|2dx+(34−a2‖uε‖2L2)∫Ω|∇^vε|2dx≤c2{‖h1‖L1+‖h2−h2,∞‖2L2},t>0, | (3.29) |
where a1, a2 and a3 are given in (3.25).
According to the uniform convergence stated in (3.2), (3.3) and (3.14), and the assumptions (1.6) and (1.7), there must exist T∗ large enough, independent of ε, such that
a3μ∫Ωuε(⋅,t)dx≤A2,t≥T∗, |
and
12∫T∗+1T∗‖uε(⋅,t)‖2L2dt+12∫T∗+1T∗‖∇^vε(⋅,t)‖2L2dt+c2∫∞T∗(‖h1‖L1+‖h2−h2,∞‖2L2)ds≤A2, |
where A:=min{18a2,18√1a1+1−18}. By using mean value theorem we can find ˆt0∈(T∗,T∗+1), depending on ε, such that
y(ˆt0)+c2∫∞ˆt0(‖h1‖L1+‖h2−h2,∞‖2L2)ds≤A, | (3.30) |
which, together with the definition of y(ˆt0), further implies that
a2‖uε(⋅,ˆt0)‖2L2≤2a2A, | (3.31) |
and
‖∇^vε(⋅,ˆt0)‖2L2≤2A. | (3.32) |
We now claim that
a2‖uε(⋅,t)‖2L2≤4a2A,t≥ˆt0, | (3.33) |
‖∇^vε(⋅,t)‖2L2≤4A,t≥ˆt0, | (3.34) |
and thereby assert
a1‖uε(⋅,t)‖2L2(‖∇^vε(⋅,t)‖2L2+1)≤4a1A(4A+1),t≥ˆt0. | (3.35) |
Indeed, the continuities of ‖∇^vε(⋅,t)‖2L2 and a2‖uε(⋅,t)‖2L2, invoking (3.31) and (3.32), show that
˜T:=sup{t|supˆt0≤s≤ta2‖uε(⋅,t)‖2L2≤4a2A,supˆt0≤s≤t‖∇^vε(⋅,t)‖2L2≤4A,}>ˆt0, | (3.36) |
and so we only need to show that ˜T=∞. If on the contrary, at least one of the following statements must hold
supˆt0≤s≤˜Ta2‖uε(⋅,t)‖2L2=4a2A, | (3.37) |
supˆt0≤s≤˜T‖∇^vε(⋅,t)‖2L2=4A, | (3.38) |
which, together with the definition of A, further leads to
a1‖uε(⋅,t)‖2L2(‖∇^vε(⋅,t)‖2L2+1)≤4a1A(4A+1)≤14,t∈[ˆt0,˜T], | (3.39) |
34−a2‖uε(⋅,t)‖2L2≥34−4a2A≥14,t∈[ˆt0,˜T]. | (3.40) |
However, it follows from (3.29), (3.39) and (3.40) that
y′(t)+c1y(t)+12∫Ω|∇uε|2dx+14∫Ω|Δ^vε|2dx≤c2{‖h1‖L1+‖h2−h2,∞‖2L2},t∈[ˆt0,˜T]. |
This, by means of the standard ODE techniques, results in that for any t∈[ˆt0,˜T]
y(t)≤e−c1(t−ˆt0)y(ˆt0)+e−c1t∫tˆt0ec1sc2{‖h1‖L1+‖h2−h2,∞‖2L2}ds≤y(ˆt0)+c2∫∞ˆt0{‖h1‖L1+‖h2−h2,∞‖2L2}ds, |
which, combined with (3.30) and the definition of y(t), implies
y(t)≤A,t∈[ˆt0,˜T]. |
Hence, recalling the definitions of y(t) and A again, we must have
a2‖uε(⋅,t)‖2L2≤2a2Aand‖∇^vε(⋅,t)‖2L2≤2A,t∈[ˆt0,˜T], |
which contradicts (3.37) and (3.38). Thus we have that ˜T=∞, and prove (3.33)–(3.35) as desired.
Based on the definition of A and the validity of (3.33)–(3.35), we see from (3.29) that
y′(t)+c1y(t)+12∫Ω|∇uε|2dx+14∫Ω|Δ^vε|2dx≤c2{‖h1‖L1+‖h2−h2,∞‖2L2},t≥ˆt0. | (3.41) |
Hence, using the standard ODE techniques again, yields that for any t≥ˆt0
y(t)≤y(ˆt0)+c2∫tˆt0{‖h1‖L1+‖h2−h2,∞‖2L2}ds, |
which, together with (3.30), ensures
y(t)≤c3,t≥ˆt0. | (3.42) |
This evidently entails (3.27).
Moreover, integrating (3.41) over [s,t] with ˆt0≤s≤t and using (1.6) and (1.7) again, we subsequently arrive at
y(t)+12∫ts∫Ω|∇uε|2dxdτ+14∫ts∫Ω|Δ^vε|2dxdτ≤y(s)+c4. |
Based on (3.42), we get that Eε(s)≤c3 due to s≥ˆt0, and thereby obtain (3.28) as desired.
In view of Lemma 3.5 and the boundedness criterion obtained in [41,42] via the Moser iteration and the semigroup theory, we can get the eventual bound of the generalized solution.
Lemma 4.1. Let T0 be given in Lemma 3.5. Then there exists a5>0, with the property that for any q>2
‖∇^vε(⋅,t)‖Lq≤a5,t≥T0+1. | (4.1) |
Proof. By means of (3.6) and the properties of the Neumann heat semigroup (cf. [43,Lemma 1.3] and [44,Lemma 2.1]), for all t>T0 and q>2 we have
‖∇^vε(⋅,t)‖Lq≤‖∇e(t−T0)(Δ−1)^vε(⋅,T0)‖Lq+∫tT0‖∇e(t−s)(Δ−1)(uεvε1+εuεvε+h2−h2,∞)‖Lqds≤c1(1+(t−T0)−(12−1q))‖∇^vε(⋅,T0)‖L2+c1∫tT0(1+(t−s)−12−(12−1q))e−(t−s)(‖uεvε‖L2+‖h2−h2,∞‖L2)ds, |
which, combined with (1.4) and (3.27), reduces to
‖∇^vε(⋅,t)‖Lq≤c2+c2(t−T0)−(12−1q)+c2∫tT0(1+(t−s)−12−(12−1q))e−(t−s)‖uεvε‖L2ds. |
An application of Hölder's inequality, invoking (3.27) and Lemma 2.4, yields that for any t≥T0
‖uεvε‖L2≤‖uε‖L2‖vε‖L∞≤c3(‖^vε‖L∞+1), |
which, with the help of the Gagliardo-Nirenberg inequality, entails
‖uεvε‖L2≤c4(‖^vε‖1−ϑL2‖∇^vε‖ϑLq+‖^vε‖L2+1), |
where ϑ:=q2(q−1). By employing (3.7), we arrive at
‖uεvε‖L2≤c5(‖∇^vε‖ϑLq+1). |
Collecting these, it follows that for any t>T0
‖∇^vε(⋅,t)‖Lq≤c6+c2(t−T0)−(12−1q)+c6∫tT0(1+(t−s)−12−(12−1q))e−(t−s)‖∇^vε‖ϑLqds. |
Letting K(T):=supt∈(T0,T)‖∇^vε(⋅,t)‖Lq for any T∈(T0,∞), we get
K(T)≤c6+c2(t−T0)−(12−1q)+c7Kϑ(T), |
which, by using Young's inequality, ensures
K(T)≤c8+c2(t−T0)−(12−1q),t>T0. |
Hence, for any t≥T0+1, we arrive at (4.1).
Based on (4.1), we can obtain the time-independent bound for uε in L∞(Ω).
Lemma 4.2. Let T0 be given in Lemma 3.5. Then there exists a6>0, such that
‖uε(⋅,t)‖L∞≤a6,t≥T0+2. | (4.2) |
Proof. From the constant variation formula associated with the first equation in (2.1), we get that for any t>t1:=T0+1
0≤uε(x,t)=e(Δ−1)(t−t1)uε(x,t1)+∫tt1e(Δ−1)(t−s)(−χ∇⋅(uε∇lnvε)−κuεvε+ruε−μu2ε+h1+uε)ds≤e(Δ−1)(t−t1)uε(x,t1)+∫tt1e(Δ−1)(t−s)(−χ∇⋅(uε∇lnvε)+ruε+h1+uε)ds, |
which, with the help of the properties of Neumann heat semigroup (cf. [43,Lemma 1.3] and [44,Lemma 2.1]), we can pick c1>0 such that
‖uε(⋅,t)‖L∞≤c1(1+(t−t1)−12)‖uε(⋅,t1)‖L2+c1∫tt1(1+(t−s)−12)e−(t−s)‖uε+h1‖L2ds+c1∫tt1(1+(t−s)−12−13)e−(t−s)‖uε∇lnvε‖L3ds. |
Hence, (3.27) and (1.4) show that
‖uε(⋅,t)‖L∞≤c2+c2(t−t1)−12+c1∫tt1(1+(t−s)−12−13)e−(t−s)‖uε∇lnvε‖L3ds. |
On the basis of Hölder's inequality and (2.4), it follows that
‖uε∇lnvε‖L3≤‖uε‖L4‖∇vε‖L12‖v−1ε‖L∞≤η−1‖uε‖L4‖∇vε‖L12, |
which, together with (4.1) and the fact that ^vε=vε−v∞, entails
‖uε∇lnvε‖L3≤c3‖uε‖L4,t≥T0+1. |
Based on this, the interpolation inequality and (3.27) indicate that
‖uε∇lnvε‖L3≤c3‖uε‖12L2‖uε‖12L∞≤c4‖uε‖12L∞,t≥T0+1. |
Collecting these, we arrive at
‖uε(⋅,t)‖L∞≤c2+c2(t−t1)−12+c5∫tt1(1+(t−s)−12−13)e−(t−s)‖uε‖12L∞ds. |
Setting K(T):=supt∈(t1,T)‖uε(⋅,t)‖L∞ for any T∈(t1,∞), we have
K(T)≤c2+c2(t−t1)−12+c6K12(T). |
Using Young's inequality yields
K(T)≤c7+c2(t−t1)−12,t>t1, |
which must lead to
K(T)≤c8,t≥t1+1. |
This implies (4.2) directly.
A straightforward consequence of Lemmas 4.1 and 4.2, invoking the parabolic Schauder estimates [45], can be stated as follows.
Lemma 4.3. There exists a7>0, independent of ε and t, with the property that for some α∈(0,1)
‖uε(⋅,s)‖C2+α,1+α2(Ω×[t,t+1])+‖vε(⋅,s)‖C2+α,1+α2(Ω×[t,t+1])≤a7,t>T0+2, | (4.3) |
where T0 is given in Lemma 3.5.
Proof. Based on Lemmas 4.1 and 4.2 and the Schauder estimates ([45]), a straightforward reasoning involving standard bootstrap techniques ensures that (4.3) holds as desired by recalling Lemma 2.4 and ^vε=vε−v∞.
Lemma 4.3, combined with the Arzelà-Ascoli theorem, is enough to prove that the generalized solution (u,v) established in [3] admits the desired regularity in Theorem 1.1.
Lemma 4.4. Let (u,v) be a generalized solution stated in Definition 1.1, and T0 be given in Lemma 3.5. Then there exists a8>0 with the property that for any q>2
‖u(⋅,t)‖L∞+‖v(⋅,t)‖W1,q+‖v−1(⋅,t)‖L∞≤a8,t≥T0+2, | (4.4) |
‖u(⋅,s)‖C2,1(Ω×[t,t+1])+‖v(⋅,s)‖C2,1(Ω×[t,t+1])≤a8,t>T0+2. | (4.5) |
Moreover, u≥0, v>0 and (u,v) solves the initial-boundary value problem (1.2) classically in Ω×(T0+2,∞).
Proof. Invoking Lemma 4.3, [3,Lemma 4.2] and the Arzelà-Ascoli theorem, there exists a subsequence of {εj}∞j=1 (still expressed as {εj}∞j=1) such that for any t>T0+2, as ε=εj→0,
uε→uinC2,1(Ω×[t,t+1],vε→vinC2,1(Ω×[t,t+1]. |
This ensures (4.3), and thereby (4.4) holds as desired by using Sobolev's inequality, (4.1), (2.4) and (4.2) again. Moreover, along the lines demonstrated in [46,Lemma 2.1], we can see that if u≥0 and v>0 satisfying (4.3) and such that (u,v) is a generalized solution of (1.3) in the sense of Definition 1.1, then (u,v) also solves (1.3) in the classical sense in Ω×(T0+2,∞).
Asymptotic behavior of the generalized solution featured in Theorem 1.1 is now almost immediate.
Lemma 5.1. Let all assumptions in Theorem 1.1 be fulfilled. Then
‖uε(⋅,t)‖L∞+‖vε(⋅,t)−v∞(⋅)‖L∞→0,ast→∞, | (5.1) |
where v∞ denotes the solution of the boundary value problem (1.15).
Proof. It directly follows from (3.13) that
∫Ω|^vε|2(⋅,t)dx→0ast→∞uniformlyinε. | (5.2) |
Using Sobolev's inequality and (4.1) again, for some r>2 there exist c2>0 and c3>0 such that
‖^vε(⋅,t)‖L∞≤c2‖^vε(⋅,t)‖r−22(r−1)L2‖^vε(⋅,t)‖r2(r−1)W1,r≤c3‖^vε(⋅,t)‖r−22(r−1)L2,t≥T0+2, |
which, in conjunction with (5.2), entails
‖^vε(⋅,t)‖L∞→0,ast→∞uniformlyinε. | (5.3) |
To get the decay on ‖uε‖L∞, we further develop the method used in [47]. According to the variation-of-constants formula for uε, for t0:=T0+2 the known estimates for the Neumann heat semigroup ensure that for any t>t0
‖uε(⋅,t)‖L∞≤‖e(Δ−1)(t−t0)uε(⋅,t0)‖L∞+χ∫tt0‖e(Δ−1)(t−s)∇⋅(uε∇lnvε)‖L∞ds+∫tt0‖e(Δ−1)(t−s)(ruε+h1+uε)‖L∞ds≤c4(1+(t−t0)−13)e−δ(t−t0)‖uε(⋅,t0)‖L3+c4∫tt0(1+(t−s)−(13+12))e−δ(t−s)‖uε∇lnvε‖L3ds+c4∫tt0(1+(t−s)−13)e−δ(t−s)‖ruε+h1+uε‖L3ds=:V1+V2+V3, |
with some c4>0 and δ>0. As a consequence of (4.4), we can find c5>0, independent of ε, such that
V1≤c5(1+(t−t0)−13)e−δ(t−t0)≤2c5et0e−t,t≥t0+1, |
which clearly implies that for fixed t0
V1→0,ast→∞uniformlyinε. | (5.4) |
For V2, Hölder's inequality, combined with (2.4), (4.4) and Lemma 2.4, entails
‖uε∇lnvε‖L3≤‖v−1ε‖L∞‖uε‖L6‖∇vε‖L6≤‖v−1ε‖L∞‖uε‖L6(‖∇ˆvε‖L6+‖∇v∞‖L6)≤c6‖uε‖L6,t≥t0. |
By further assuming that t>2t0 and letting
V21=c4c6∫t2t0(1+(t−s)−56)e−δ(t−s)‖uε‖L6ds,V22=c4c6∫tt2(1+(t−s)−56)e−δ(t−s)‖uε‖L6ds, |
it follows that
V2≤V21+V22,t>2t0. |
For V21, using (4.4) again we have
V21=c4c6∫(t−t0)t2(1+s−56)e−δs‖uε‖L6ds≤c7∫tt2(1+s−56)e−δsds. |
Due to the fact that
∫∞0(1+s−56)e−δsds≤c8, |
we infer that
V21→0,ast→∞uniformlyinε. | (5.5) |
For V22, an application of the interpolation inequality and (4.4) yields that
‖uε(⋅,t)‖L6≤‖uε(⋅,t)‖16L1‖uε(⋅,t)‖56L∞≤c9‖uε(⋅,t)‖16L1,t>2t0. |
Invoking this, we arrive at
V22≤c10sups>t2‖uε(⋅,s)‖16L1∫tt2(1+(t−s)−56)e−δ(t−s)ds≤c11sups>t2‖uε(⋅,s)‖16L1,t>2t0, |
which, combined with (3.2), entails
V22→0,ast→∞uniformlyinε. |
This, together with (5.5), implies
V2→0,ast→∞uniformlyinε. | (5.6) |
Similarity, we set
V31=c4∫t2t0(1+(t−s)−13)e−δ(t−s)‖(r+1)uε+h1‖L3ds,V32=c4∫tt2(1+(t−s)−13)e−δ(t−s)‖(r+1)uε+h1‖L3ds, |
and thereby get
V3≤V31+V32,t>2t0. |
Similar to (5.5), we can infer from (4.4), (1.4) and Hölder's inequality that
V31≤c12∫tt2(1+s−13)e−δsds,t>2t0, |
and hence
V31→0,ast→∞uniformlyinε. | (5.7) |
Similar to the estimate for V22, it follows from the interpolation inequality and (4.4) that
c4∫tt2(1+(t−s)−13)e−δ(t−s)‖(r+1)uε‖L3ds≤c13sups>t2‖uε(⋅,s)‖13L1,t>2t0, |
which, combined with (3.2), entails
c4∫tt2(1+(t−s)−13)e−δ(t−s)‖(r+1)uε‖L3ds→0,ast→∞uniformlyinε. |
On the other hand, the interpolation inequality and (1.4) imply
‖h1‖L3≤‖h1‖13L1‖h1‖23L∞≤c14‖h1‖13L1. |
This, with the help of Hölder's inequality, ensures
c4∫tt2(1+(t−s)−13)e−δ(t−s)‖h1(⋅,s)‖L3ds≤c15{∫tt2(1+(t−s)−12)e−32δ(t−s)ds}23{∫tt2‖h1(⋅,s)‖3L3ds}13≤c16{∫tt2‖h1(⋅,s)‖L1ds}13,t>2t0, |
which, together with (1.6), leads to
c4∫tt2(1+(t−s)−13)e−δ(t−s)‖h1(⋅,s)‖L3ds→0,ast→∞uniformlyinε. |
Hence, we arrive at
V32→0,ast→∞uniformlyinε, |
which, in conjunction with (5.7), gives us
V3→0,ast→∞uniformlyinε. |
This, further combined with (5.4) and (5.6), asserts
‖uε(⋅,t)‖L∞→0,ast→∞uniformlyinε, |
which implies that (5.1) holds as desired by recalling (5.3) and the definition of ^vε.
Our main result on eventual smoothness and stabilization in Theorem 1.1 is in fact a by-product of our previous analysis.
Proof of Theorem 1.1. The eventual smoothness in Theorem 1.1 has been verified evidently in Lemma 4.4. For the stabilization, it readily follows from Lemma 5.1, Lemma 4.3, [3,Lemma 4.2] and the Arzelà-Ascoli theorem that (1.14) holds.
The authors are very grateful to the referees for their detailed comments and valuable suggestions, which greatly improved the manuscript. The research of BL is supported by the Natural Science Foundation of Ningbo Municipality (No. 2022J147).
The authors declare there is no conflicts of interest.
[1] |
M. Short, M. Drsogna, V. Pasour, G. Tita, P. Brantingham, A. Bertozzi, et al., A statistical model of criminal behavior, Math. Models Methods Appl. Sci., 18 (2008), 1249–1267. https://doi.org/10.1142/S0218202508003029 doi: 10.1142/S0218202508003029
![]() |
[2] |
M. Short, A. Bertozzi, P. Brantingham, G. Tita, Dissipation and displacement of hotspots in reaction-diffusion model of crime, Proc. Natl. Acad. Sci. USA, 107 (2010), 3961–3965. https://doi.org/0.1073/pnas.0910921107 doi: 10.1073/pnas.0910921107
![]() |
[3] |
F. Heihoff, Generalized solutions for a system of partial differential equations arising from urban crime modeling with a logistic source term, Z. Für Angew. Math. Phys., 71 (2020), 80. https://doi.org/10.1007/s00033-020-01304-w doi: 10.1007/s00033-020-01304-w
![]() |
[4] |
N. Bellomo, F. Colasuonno, D. Knopoff, J. Soler, From a systems theory of sociology to modeling the onset and evolution of criminality, Networks Heterog. Media, 10 (2015), 421–441. https://doi.org/10.3934/nhm.2015.10.421 doi: 10.3934/nhm.2015.10.421
![]() |
[5] |
H. Berestycki, J. Nadal, Self-organised critical hot spots of criminal activity, Eur. J. Appl. Math., 21 (2010), 371–399. https://doi.org/10.1017/S0956792510000185 doi: 10.1017/S0956792510000185
![]() |
[6] |
Y. Gu, Q. Wang, G. Yi, Stationary patterns and their selection mechanism of urban crime models with heterogeneous near-repeat victimization effect, Eur. J. Appl. Math., 28 (2017), 141–178. https://doi.org/10.1017/S0956792516000206 doi: 10.1017/S0956792516000206
![]() |
[7] |
A. Pitcher, Adding police to a mathematical model of burglary, Eur. J. Appl. Math., 21 (2010), 401–419. https://doi.org/10.1017/S0956792510000112 doi: 10.1017/S0956792510000112
![]() |
[8] |
M. Short, G. Mohler, P. Brantingham, G. Tita, Gang rivalry dynamics via coupled point process networks, Discrete Contin. Dyn. Syst. Ser. B, 19 (2014), 1459–1477. https://doi.org/10.3934/dcdsb.2014.19.1459 doi: 10.3934/dcdsb.2014.19.1459
![]() |
[9] |
W. Tse, M. Ward, Asynchronous instabilities of crime hotspots for a 1-D reaction-diffusion model of urban crime with focused police patrol, SIAM J. Appl. Dyn. Syst., 17 (2018), 2018–2075. https://doi.org/10.1137/17M1162585 doi: 10.1137/17M1162585
![]() |
[10] |
J. Zipkin, M. Short, A. Bertozzi, Cops on the dots in a mathematical model of urban crime and police response, Discrete Contin. Dyn. Syst. Ser. B, 19 (2014), 1479–1506. https://doi.org/10.3934/dcdsb.2014.19.1479 doi: 10.3934/dcdsb.2014.19.1479
![]() |
[11] |
N. Bellomo, N. Outada, J. Soler, Y. Tao, M. Winkler, Chemotaxis and cross-diffusion models in complex environments: Models and analytic problems toward a multiscale vision, Math. Models Methods Appl. Sci., 32 (2022), 713–792. https://doi.org/10.1142/S0218202522500166 doi: 10.1142/S0218202522500166
![]() |
[12] |
M. D'Orsogna, M. Perc, Statistical physics of crime: a review, Phys. Life Rev., 12 (2015), 1–21. https://doi.org/10.1016/j.plrev.2014.11.001 doi: 10.1016/j.plrev.2014.11.001
![]() |
[13] |
N. Rodríguez, A. Bertozzi, Local existence and uniqueness of solutions to a PDE model for criminal behavior, Math. Models Methods Appl. Sci., 20 (2010), 1425–1457. https://doi.org/10.1142/S0218202510004696 doi: 10.1142/S0218202510004696
![]() |
[14] |
N. Rodríguez, M. Winkler, On the global existence and qualitative behavior of one-dimensional solutions to a model for urban crime, Eur. J. Appl. Math., 33 (2022), 919–959. https://doi.org/10.1017/S0956792521000279 doi: 10.1017/S0956792521000279
![]() |
[15] |
Q. Wang, D. Wang, Y. Feng, Global well-posedness and uniform boundedness of urban crime models: One-dimensional case, J. Differ. Equations, 269 (2020), 6216–6235. https://doi.org/10.1016/j.jde.2020.04.035 doi: 10.1016/j.jde.2020.04.035
![]() |
[16] |
M. Freitag, Global solutions to a higher-dimensional system related to crime modeling, Math. Meth. Appl. Sci., 41 (2018), 6326–6335. https://doi.org/10.1002/mma.5141 doi: 10.1002/mma.5141
![]() |
[17] |
J. Shen, B. Li, Mathematical analysis of a continuous version of statistical models for criminal behavior, Math. Meth. Appl. Sci., 43 (2020), 409–426. https://doi.org/10.1002/mma.5898 doi: 10.1002/mma.5898
![]() |
[18] |
J. Ahn, K. Kang, J. Lee, Global well-posedness of logarithmic Keller-Segel type systems, J. Differ. Equations, 287 (2021), 185–211. https://doi.org/10.1016/j.jde.2021.03.053 doi: 10.1016/j.jde.2021.03.053
![]() |
[19] |
Y. Tao, M. Winkler, Global smooth solutions in a two-dimensional cross-diffusion system modeling propagation of urban crime, Commun. Math. Sci., 19 (2021), 829–849. https://doi.org/10.4310/CMS.2021.v19.n3.a12 doi: 10.4310/CMS.2021.v19.n3.a12
![]() |
[20] |
M. Winkler, Global solvability and stabilization in a two-dimensional cross-diffusion system modeling urban crime propagation, Ann. Inst. Henri Poincaré, Anal. Non Linéaire, 36 (2019), 1747–1790. https://doi.org/10.1016/j.anihpc.2019.02.004 doi: 10.1016/j.anihpc.2019.02.004
![]() |
[21] |
Y. Jiang, L. Yang, Global solvability and stabilization in a three-dimensional cross-diffusion system modeling urban crime propagation, Acta Appl. Math., 178 (2022), 11. https://doi.org/10.1007/s10440-022-00484-z doi: 10.1007/s10440-022-00484-z
![]() |
[22] | B. Li, L. Xie, Generalized solution to a 2D parabolic-parabolic chemotaxis system for urban crime: Global existence and large time behavior, submitted for publication, 2022. |
[23] |
N. Rodríguez, M. Winkler, Relaxation by nonlinear diffusion enhancement in a two-dimensional cross-diffusion model for urban crime propagation, Math. Models Methods Appl. Sci., 30 (2020), 2105–2137. https://doi.org/10.1142/S0218202520500396 doi: 10.1142/S0218202520500396
![]() |
[24] |
L. Yang, X. Yang, Global existence in a two-dimensional nonlinear diffusion model for urban crime propagation, Nonliear Anal., 224 (2022), 113086. https://doi.org/10.1016/j.na.2022.113086 doi: 10.1016/j.na.2022.113086
![]() |
[25] | M. Fuest, F. Heihoff, Unboundedness phenomenon in a reduced model of urban crime, preprint, arXiv: 2109.01016. |
[26] | B. Li, L. Xie, Global large-data generalized solutions to a two-dimensional chemotaxis system stemming from crime modelling, Discrete Contin. Dyn. Syst. Ser. B, 2022. https://doi.org/10.3934/dcdsb.2022167 |
[27] |
B. Li, Z. Wang, L. Xie, Regularization effect of the mixed-type damping in a higher-dimensional logarithmic Keller-Segel system related to crime modeling, Math. Biosci. Eng., 24 (2023), 4532–4559. https://doi.org/10.3934/mbe.2023210 doi: 10.3934/mbe.2023210
![]() |
[28] |
R. Manásevich, Q. Phan, P. Souplet, Global existence of solutions for a chemotaxis-type system arising in crime modelling, Eur. J. Appl. Math., 24 (2013), 273–296. https://doi.org/10.1017/S095679251200040X doi: 10.1017/S095679251200040X
![]() |
[29] |
N. Rodríguez, On the global well-posedness theory for a class of PDE models for criminal activity, Phys. D Nonlinear Phenom., 260 (2013), 191–200. https://doi.org/10.1016/j.physd.2012.08.003 doi: 10.1016/j.physd.2012.08.003
![]() |
[30] |
D. Wang, Y. Feng, Global well-posedness and uniform boundedness of a higher dimensional crime model with a logistic source term, Math. Meth. Appl. Sci., 45 (2022), 4727–4740. https://doi.org/10.1002/mma.8066 doi: 10.1002/mma.8066
![]() |
[31] |
T. Black, Eventual smoothness of generalized solutions to a singular chemotaxis-Stokes system in 2D, J. Differ. Equations, 265 (2018), 2296–2339. https://doi.org/10.1016/j.jde.2018.04.035 doi: 10.1016/j.jde.2018.04.035
![]() |
[32] |
T. Black, C. Wu, Prescribed signal concentration on the boundary: eventual smoothness in a chemotaxis-Navier-Stokes system with logistic proliferation, Calc. Var., 61 (2022), 96. https://doi.org/10.1007/s00526-022-02201-y doi: 10.1007/s00526-022-02201-y
![]() |
[33] |
M. Ding, J. Lankeit, Generalized solutions to a chemotaxis-Navier-Stokes system with arbitrary superlinear degradation, SIAM J. Math. Anal., 54 (2022), 1022–1052. https://doi.org/10.1137/21M140907X doi: 10.1137/21M140907X
![]() |
[34] |
Y. Tao, M. Winkler, Eventual smoothness and stabilization of large-data solutions in a three-dimensional chemotaxis system with consumption of chemoattractant, J. Differ. Equations, 252 (2012), 2520–2543. https://doi.org/10.1016/j.jde.2011.07.010 doi: 10.1016/j.jde.2011.07.010
![]() |
[35] | B. Li, L. Xie, Generalized solution and eventual smoothness in a logarithmic Keller-Segel system for criminal activities, Math. Models Methods Appl. Sci., 2023. https://doi.org/10.1142/S0218202523500306 |
[36] |
M. Aida, K. Osaka, T. Tsujikawa, M. Mimura, Chemotaxis and growth system with sigular sensitivity function, Nonliear Anal. Real Word Appl., 6 (2005), 323–336. https://doi.org/10.1016/j.nonrwa.2004.08.011 doi: 10.1016/j.nonrwa.2004.08.011
![]() |
[37] |
M. Winkler, Global solutions in a fully parabolic chemotaxis system with singular sensitivity, Math. Methods Appl. Sci., 34 (2011), 176–190. https://doi.org/10.1002/mma.1346 doi: 10.1002/mma.1346
![]() |
[38] |
T. Xiang, Finite time blow-up in the higher dimensional parabolic-elliptic-ODE minimal chemotaxis-haptotaxis system, J. Differ. Equations, 336 (2022), 44–72. https://doi.org/10.1016/j.jde.2022.07.015 doi: 10.1016/j.jde.2022.07.015
![]() |
[39] |
T. Hillen, K. Painter, M. Winkler, Convergence of a cancer invasion model to a logistic chemotaxis model, Math. Models Methods Appl. Sci., 23 (2013), 165–198. https://doi.org/10.1142/S0218202512500480 doi: 10.1142/S0218202512500480
![]() |
[40] | O. Ladyzhenskaya, N. Ural'tseva, Linear and Quasilinear Elliptic Equations, Translated from the Russian by Scripta Technica, Inc. Translation editor: Leon Ehrenpreis, Academic Press, New York-London, 1968. |
[41] |
N. Bellomo, A. Bellouquid, Y. Tao, M. Winkler, Toward a mathematical theory of Keller-Segel models of pattern formation in biological tissues, Math. Models Methods Appl. Sci., 25 (2015), 1663–1763. https://doi.org/10.1142/S021820251550044X doi: 10.1142/S021820251550044X
![]() |
[42] |
T. Xiang, Boundedness and global existence in the higher-dimensional parabolic-parabolic chemotaxis system with/with-out growth source, J. Differ. Equations, 258 (2015), 4275–4323. https://doi.org/10.1016/j.jde.2015.01.032 doi: 10.1016/j.jde.2015.01.032
![]() |
[43] |
M. Winkler, Aggregation vs. global diffusive behavior in the higher-dimensional Keller-Segel model, J. Differ. Equations, 248 (2010), 2889–2905. https://doi.org/10.1016/j.jde.2010.02.008 doi: 10.1016/j.jde.2010.02.008
![]() |
[44] |
X. Cao, Global bounded solutions of the higher-dimensional Keller-Segel system under smallness conditions in optimal spaces, Discrete Contin. Dyn. Syst., 35 (2015), 1891–1904. https://doi.org/10.3934/dcds.2015.35.1891 doi: 10.3934/dcds.2015.35.1891
![]() |
[45] | O. Ladyzhenskaya, V. Solonnikov, N. Ural'tseva, Linear and Quasilinear Equations of Parabolic Type, Translated from the Russian by S. Smith. Translations of Mathematical Monographs, American Mathematical Society, Providence, R.I., 1968. |
[46] |
M. Winkler, Large-data global generalized solutions in a chemotaxis system with tensor-valued sensitivities, SIAM J. Math. Anal., 47 (2015), 3092–3115. https://doi.org/dx.doi.org/10.1137/140979708 doi: 10.1137/140979708
![]() |
[47] |
K. Fujie, A. Ito, M. Winkler, T. Yokota, Stabilization in a chemotaxis model for tumor invasion, Discrete Contin. Dyn. Syst., 36 (2016), 151–169. https://doi.org/10.3934/dcds.2016.36.151 doi: 10.3934/dcds.2016.36.151
![]() |
1. | Bin Li, Li Xie, Can Dirac-type singularities in Keller-Segel systems be ruled out by power-type singular sensitivities?, 2024, 379, 00220396, 413, 10.1016/j.jde.2023.10.013 | |
2. | Bin Li, Li Xie, Smoothness effects of a quadratic damping term of mixed type on a chemotaxis-type system modeling propagation of urban crime, 2023, 73, 14681218, 103912, 10.1016/j.nonrwa.2023.103912 | |
3. | Bin Li, Li Xie, A chemotaxis system with singular sensitivity for burglaries in the higher-dimensional settings: generalized solvability and long-time behavior, 2024, 63, 0944-2669, 10.1007/s00526-023-02657-6 | |
4. | Bin Li, Li Xie, Generalised solution to a 2D parabolic-parabolic chemotaxis system for urban crime: Global existence and large-time behaviour, 2024, 35, 0956-7925, 409, 10.1017/S0956792523000268 | |
5. | Wenjing Jiang, Qi Wang, Global well-posedness and uniform boundedness of 2D urban crime models with nonlinear advection enhancement, 2024, 0956-7925, 1, 10.1017/S0956792524000792 |